Identifying differentially expressed peptides in global LC-MS/MS data

From Master Projects
Jump to: navigation, search


has title::Identifying differentially expressed peptides in global LC-MS/MS data
status: finished
Master: project within::Bioinformatics
Student name: student name::Michiel van Ooijen
Dates
Start start date:=2015/03/01
End end date:=2015/09/15
Supervision
Supervisor: Henk-Jan van den Ham
Second supervisor: Sanne Abeln
Company: has company::Erasmus MC
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:Posternaam.pdf

Signature supervisor



..................................

Abstract

Changes in protein levels expressed by a cell lead to changes in cell function. LC-MS/MS proteomics can monitor changes in protein levels, for instance after applying different stimuli to cells. Technical variability obscures the biological signal and combining multiple LC-MS/MS runs leads to missing values. Consequently, statistically inferring changes in protein levels from LC-MS/MS data is challenging. Published global LC-MS/MS datasets contain limited information on peptides which are truly differentially expressed. The use of simulated LC-MS/MS data can overcome this limitation. Four different methods are applied to statistically infer differentially expressed peptides in LC-MS/MS data: Two-sample T-test, Limma’s eBayes, GLM-Gamma and MSstats. Statistical tests are validated on simulated LC-MS/MS data. LC-MS/MS data is simulated using a resampling-based method and an ab initio method. The fraction of missing values and number of biological replicates present in LC-MS/MS datasets, limits the identification of differentially expressed peptides. With less than six biological replicates available, GLM-Gamma achieves the highest sensitivity. Limma’s eBayes can achieve a higher sensitivity than GLM-Gamma when at least six biological replicates are available. The fraction false positives are conservative for all methods when at least three biological replicates are available.